############################################
###### Replication Code for Figure A4 ######
########### January 9th, 2025 ##############
############################################

### Script produces Figure A4 and saves it in the `/figures` folder

rm(list=ls())

############ Balancing on predictors ################

# load library
library(dplyr)
library(fixest)
library(ggplot2)
library(stringr)
library(MatchIt)
library(ebal)
library(hbal)

# load data
df <- readRDS("data/workfile.rds")
w <- readRDS("data/weighted_party_feeling.rds")

# scale outfeeling variables
df$out_feeling <- scale(df$mean_outparty_feeling) + scale(df$mean_close_num)

# join weights to data
df <- df %>% left_join(w)

# specify control variables
control_variables <- c('political_interest_num',
                       "high_edu", "medium_edu", "low_edu", 
                       "high_income", "medium_income", "low_income", 
                       "dem_rural", "dem_age", "mean_close_num", "mean_outparty_feeling",
                       "lrscale",  "dem_female", 
                       "parfam_eco", "parfam_nat", "parfam_liberal",
                       "parfam_cd", "parfam_sd", "parfam_left", 
                       "parfam_con", 
                       "coDE", "coFR", "coES", "coNL", "coUS", "coPL", 
                       "coGR", "coIT", "coSE") 

df$trust_in_opposing_parties <- scale(df$trust_in_opposing_parties)

coef_df <- data.frame()

########## Select variables
df <- df %>% 
  select(open_immig, dem_female, dem_country_code, dem_education_level,
         dem_country_code, dem_rural, dem_age, lrscale, parfam_name,
         high_pol_interest, mean_close_num, d_affiliated,
         high_edu, medium_edu, low_edu, treated, mean_outparty_feeling, 
         political_interest_num, treated_econ, treated_culture, out_feeling,
         trust_in_opposing_parties, 
         low_income, medium_income, high_income) %>% 
  mutate(coDE = ifelse(dem_country_code == "DE", 1, 0), 
         coFR = ifelse(dem_country_code == "FR", 1, 0), 
         coNL = ifelse(dem_country_code == "NL", 1, 0),
         coUS = ifelse(dem_country_code == "US", 1, 0),
         coPL = ifelse(dem_country_code == "PL", 1, 0),
         coGR = ifelse(dem_country_code == "GR", 1, 0), 
         coIT = ifelse(dem_country_code == "IT", 1, 0), 
         coSE = ifelse(dem_country_code == "SE", 1, 0), 
         coGB = ifelse(dem_country_code == "GB", 1, 0), 
         coES = ifelse(dem_country_code == "ES", 1, 0), 
         parfam_liberal = ifelse(parfam_name == "Liberal parties", 1, 0), 
         parfam_other = ifelse(parfam_name == "Other parties", 1, 0), 
         parfam_eco = ifelse(parfam_name == "Ecological parties", 1, 0), 
         parfam_nat = ifelse(parfam_name == "Nationalist parties", 1, 0), 
         parfam_left = ifelse(parfam_name == "Socialist/left parties", 1, 0), 
         parfam_cd = ifelse(parfam_name == "Christian democratic parties", 1, 0), 
         parfam_nonvoters = ifelse(parfam_name == "Non-voters", 1, 0), 
         parfam_sd = ifelse(parfam_name == "Social democratic parties", 1, 0), 
         parfam_con = ifelse(parfam_name == "Conservative parties", 1, 0), 
         parfam_ethnic = ifelse(parfam_name == "Ethnic and regional parties", 1, 0),
         parfam_spec = ifelse(parfam_name == "Special issue party", 1, 0),)


##### Immigration - Cultural condition Left
nomiss <- df %>% 
  filter((open_immig == 0 & treated_culture == 1) | treated == 0 & lrscale < 5) %>% 
  .[complete.cases(.), ]

hbal.out <- hbal(Treat = 'treated_culture', 
                 X = control_variables,  
                 Y = 'trust_in_opposing_parties',  data = nomiss)
m1 = att(hbal.out, dr = FALSE)

temp <- data.frame(coef = c(m1$Estimate[1]),
                   se = c(m1$`Std. Error`[1]), 
                   name = "Cultural condition",
                   var = "Trust", 
                   group = "Left-wing respondents")


coef_df <- rbind.data.frame(coef_df, temp)


##### Immigration - Culturalcondition Right
nomiss <- df %>% 
  filter((open_immig == 1 & treated_culture == 1) | treated == 0 & lrscale > 5) %>% 
  .[complete.cases(.), ]

hbal.out <- hbal(Treat = 'treated', 
                 X = control_variables,  
                 Y = 'trust_in_opposing_parties',  data = nomiss)
m1 = att(hbal.out, dr = FALSE)


temp <- data.frame(coef = c(m1$Estimate[1]),
                   se = c(m1$`Std. Error`[1]), 
                   name = "Cultural condition",
                   var = "Trust", 
                   group = "Right-wing respondents")

coef_df <- rbind.data.frame(coef_df, temp)


##### Immigration - Economiccondition Right
nomiss <- df %>% 
  filter((open_immig == 1 & treated_econ == 1) | treated == 0 & lrscale > 5) %>% 
  .[complete.cases(.), ]

hbal.out <- hbal(Treat = 'treated', 
                 X = control_variables,  
                 Y = 'trust_in_opposing_parties',  data = nomiss)
m1 = att(hbal.out, dr = FALSE)


temp <- data.frame(coef = c(m1$Estimate[1]),
                   se = c(m1$`Std. Error`[1]), 
                   name = "Economic condition",
                   var = "Trust", 
                   group = "Right-wing respondents")

coef_df <- rbind.data.frame(coef_df, temp)


##### Immigration - Economiccondition Left
nomiss <- df %>% 
  filter((open_immig == 1 & treated_econ == 1) | treated == 0 & lrscale < 5) %>% 
  .[complete.cases(.), ]

hbal.out <- hbal(Treat = 'treated', 
                 X = control_variables,  
                 Y = 'trust_in_opposing_parties',  data = nomiss)
m1 = att(hbal.out, dr = FALSE)


temp <- data.frame(coef = c(m1$Estimate[1]),
                   se = c(m1$`Std. Error`[1]), 
                   name = "Economic condition",
                   var = "Trust", 
                   group = "Left-wing respondents")

coef_df <- rbind.data.frame(coef_df, temp)


### Generate figure

pd <- position_dodge(0.5)

coef_df <- coef_df %>% filter(var == "Trust")
coef_df1 = coef_df

regs <- ggplot(coef_df, aes(x = name,  y = coef,  color = group)) + 
  geom_point(position = pd, shape = 21, fill = 'white') + 
  geom_errorbar(aes( ymin = coef - se*1.959964, ymax = coef + se*1.959964), width = 0, position = pd) + 
  geom_errorbar(aes( ymin = coef - se*1.644854, ymax = coef + se*1.644854), width = 0, position = pd, linewidth = 1.2) + 
  geom_hline(yintercept = 0, color = "red", linetype = "dashed") +
  scale_colour_manual(values = c("black", "grey")) + 
  xlab("") + 
  ylab("Effect of discussing immigration on out-partisan trust") +
  coord_flip() +
  #facet_wrap(~ var) + 
  theme_bw() + 
  guides(color=guide_legend(title="")) + 
  theme(legend.position="bottom")

regs
ggsave("figures/FigA4_entropy_balancing_predictors_by_lr_fullset.pdf",
       height = 4, width = 6)